00:00:00
Hey everybody, welcome back
to Chip Stock Investor.
00:00:04
We're gonna be talking about NVIDIA today.
00:00:06
Uh, yes, we are.
00:00:08
This is too much energy for me,
especially when it's coming from you.
00:00:13
It's a little next level.
00:00:15
Let's just start with the key
points after the Q4 earnings update.
00:00:19
Are there issues with NVIDIA's business
model during this Blackwell ramp up?
00:00:24
It's still early on in the
accelerated computing or AI industry.
00:00:28
We'll talk more about that
and why that's the case.
00:00:31
And is it time to sell any
of our NVIDIA position?
00:00:36
Yeah, it's hard to believe it's
been almost a year since we did that
00:00:38
video, time to sell NVIDIA stock.
00:00:41
We'll review some of those
important points from that video.
00:00:46
So if you missed it, we did an extensive
video on I. P. in the semiconductor
00:00:52
industry about a month ago, and we did
talk about NVIDIA and how they have been
00:00:57
able to monetize the data center, which
is very quickly becoming the largest end
00:01:02
market for semiconductors, for chips,
not just in their core competency, GPUs
00:01:08
or accelerated computing, but across
the entire industry supply chain.
00:01:13
It's really been an incredible story.
00:01:15
We have early estimates from all of the
hyperscalers, the big cloud providers,
00:01:20
that's Amazon AWS, Microsoft Azure,
Google Cloud, Oracle Cloud, plus Meta.
00:01:27
Those are the primary
customers for NVIDIA.
00:01:29
It looks like that trend is still very
much going in high gear this year in
00:01:34
2025, so things are looking good for
NVIDIA, at least from that standpoint.
00:01:41
But before we get to the actual
revenue breakdown for the data
00:01:45
center segment, let's talk about the
sponsor of today's video, public.
00:01:50
com.
00:01:51
That's right, if you are serious
about investing, you need to
00:01:54
know about public dot com.
00:01:56
That's where you can invest in everything.
00:01:58
Stocks, options, bonds, even crypto.
00:02:01
you can even earn some of the
highest yields in the industry,
00:02:04
like the 6 percent or higher yield.
00:02:07
You can lock in with the
new bond account at public.
00:02:11
Yeah, that could actually be really
significant because a lot of investors
00:02:15
still feeling like interest rates
could be headed lower later this year.
00:02:20
So even if rates do start to come
down, you do get to lock in your
00:02:24
rate at that 6 percent or higher.
00:02:26
It could be a great long term deal if
you're looking for some fixed income.c
00:02:29
Public is a FINRA registered, SIPC
insured platform that takes your
00:02:34
investments as seriously as you do.
00:02:37
Fund your account in five
minutes or less at public.
00:02:40
com forward slash CSI and
get up to 10, 000 when you
00:02:44
transfer in your old portfolio.
00:02:47
That's public.
00:02:47
com forward slash CSI.
00:02:50
This ad slot paid for by Public Investing.
00:02:53
Full disclosures in the video description.
00:02:56
Let's get back to our discussion
on NVIDIA's data center revenue.
00:02:59
Here's the revenue breakdown
for NVIDIA since 2017.
00:03:04
That giant green bar that you
see is data center revenue.
00:03:08
All the way back in 2017, that's when
Google introduced the Transformer
00:03:12
Neural Network Architecture.
00:03:14
That was the start of something big,
but it's really been about two years
00:03:19
since the NVIDIA data center inflection
point, you can see that huge growth
00:03:23
starting in 2023 and it's exponential now.
00:03:27
This all culminates in $115
billion spent in 2024, fiscal
00:03:32
year 2025 for data center alone.
00:03:35
And this is where we promised
we would point out a couple
00:03:39
of issues to be aware of.
00:03:40
Not new issues by any
stretch of the imagination.
00:03:43
We've been talking about
this for some time.
00:03:45
But with booming growth, like Nvidia
has been enjoying, there also comes
00:03:50
some very lofty expectations, and
especially when there is so much
00:03:54
riding on the data center segment now.
00:03:57
So this chart that we put
together with FinChat.
00:04:01
io, the green line tracks
NVIDIA's data center sales as a
00:04:05
percentage of total revenue, so 90
percent as we enter the new year.
00:04:10
And you may know that we like to track
what we call NVIDIA total PC revenue.
00:04:14
So we made this custom metric
with FinChat by adding together
00:04:19
NVIDIA's gaming revenue segment
and the professional visualization.
00:04:24
And this just a few years ago,
obviously used to be NVIDIA's main
00:04:28
breadwinner, and now just approaching
about 10 percent of total sales.
00:04:33
If you find custom metrics like this
helpful, you can check that out on
00:04:37
Fin Chat dot I O forward slash CSI.
00:04:40
You can get your own pro subscription
and make custom metrics just like this.
00:04:45
But Kasey, the point here is a data center
is what is moving this business right now,
00:04:50
and so whenever you have some change in
the financial assumptions for the data
00:04:57
center segment, it could wreak some havoc
on ultimately what investors are expecting
00:05:02
and therefore also the stock price, which
is probably why so many of you are here.
00:05:07
What's NVIDIA's stock price going to do?
00:05:09
Here's the outlook that NVIDIA provided
for the first quarter of fiscal year 2026.
00:05:15
They expect the revenue to be 43 billion,
which would imply a 65 percent year
00:05:20
over year revenue growth, but only
9 percent quarter over quarter, and
00:05:24
in red, we have highlighted that the
gross margins are expected to be 70.
00:05:28
6 on a GAAP basis and 71
percent on a non GAAP basis.
00:05:34
This is where we want to
key in on that first point.
00:05:37
Is there a problem with the business
model during this Blackwell ramp up?
00:05:41
You can get a hint why
that this is an issue.
00:05:43
So you said Kasey revenue expected
to be up 65 percent year over year,
00:05:49
but GAAP and non GAAP net income only
expected to be up 42 percent and 49%.
00:05:57
That implies something is going
on with the profit margin.
00:06:00
Profit margin is contracting
even as revenue increases.
00:06:04
With any new product ramp up we would
expect to see some dip in gross margin
00:06:08
that just makes sense when manufacturing
a new product, there's new equipment to
00:06:13
be installed, new processes to get worked
out . With that Blackwell ramp up, you
00:06:19
can see the decrease in gross margins.
00:06:22
Currently at 73 percent and they're
expecting it to drop another few
00:06:27
percentages in Q1 Now, with that
said, they do expect that gross
00:06:33
margins will return to the mid 70
percent range later in the year.
00:06:38
But when exactly will that happen?
00:06:40
That's a question analysts asked
multiple times on the Q and A
00:06:44
session during the Q4 report.
00:06:47
When will we see a return to
that mid 70 percent range?
00:06:50
Is it going to be in Q2?
00:06:52
Is it Q3, Q4 this year?
00:06:54
Because in addition to Blackwell currently
growing its sales base and dragging the
00:06:59
gross margin down, we also have Blackwell
Ultra that will enter production the
00:07:04
second half of 2025, calendar year 2025.
00:07:08
Is that also going to be a drag on gross
margins when that enters production?
00:07:12
And then after that we have Rubin and
then Jensen Huang said we'll be getting a
00:07:16
hint of the next computing platform at GTC
here later in March in just a few weeks.
00:07:23
This begs the question, Kasey,
is mid 70 percent gross margin
00:07:28
even a good expectation anymore
for NVIDIA going forward?
00:07:32
You're absolutely right.
00:07:33
Of course, Jensen did make some comments
regarding this ramp up and the shift
00:07:39
to each different computing platform.
00:07:41
Sometimes it's not a complete overhaul,
so to speak, and so the ramp up is not
00:07:45
maybe as aggressive in some instances.
00:07:48
In addition to the shift in
computing platforms, they
00:07:51
also have Spectrum X Ethernet.
00:07:54
Yeah, there's really two different
parts to NVIDIA's networking business,
00:07:58
which is reported within the data
center segment Quantum X is InfiniBand.
00:08:03
That's the technology they got
when they acquired Mellanox.
00:08:06
And let's just say for all intents
and purposes, this is getting degraded
00:08:10
and going away in lieu of the newer
SpectrumX Ethernet business, which
00:08:18
NVIDIA very much has some competition.
00:08:20
That's why we've been talking
about Arista Networks.
00:08:22
We had the video last
Friday on that as well.
00:08:25
And so it's not just Blackwell alone
that's a ramp up issue for margins.
00:08:30
You can see in Q4, the 9 percent year
over year decline in the networking
00:08:36
revenue part of data center.
00:08:38
They do expect this to also
return to growth starting in
00:08:42
Q1 of the new fiscal year.
00:08:44
So that's the good news.
00:08:45
But as they make this
transition to Spectrum X,
00:08:49
moving away from NVLink 8 with InfiniBand
and moving to this newer platform, NVLink
00:08:53
72 with Spectrum X Ethernet networking, we
think this is another reason why perhaps
00:09:01
the gross margin is under pressure,
And it's perhaps time to start thinking
00:09:07
about NVIDIA gross margins normalizing
somewhere in the low 70 percent range,
00:09:12
versus the mid 75 percent range.
00:09:15
Again, not a new issue.
00:09:17
This has been coming up now for some time.
00:09:19
How sustainable are those mid to high
70 percent gross margins, an issue to be
00:09:23
aware of this year because that will be a
headwind for NVIDIA's profitable growth,
00:09:30
the actual bottom line, even as revenue
continues to grow at a very fast pace.
00:09:34
Let's move on to our second key point.
00:09:37
Where are we in the accelerated
computing or AI industry?
00:09:41
Are we still early on in this industry?
00:09:44
Even though it seems like things
have changed rapidly, we actually
00:09:47
still have a long runway.
00:09:50
Nick talked about this in a
recent video that we'll link right
00:09:53
here regarding Arista Networks,
00:09:55
talking about pre training, post training,
and test time scaling, or long thinking.
00:10:01
Getting a computing system to get to
that point, long thinking, is actually
00:10:06
going to be a bit of a process.
00:10:08
Yeah, and each of these step
ups in, AI scaling laws requires
00:10:14
more computing power, not less.
00:10:16
And this was the point with
Deep seek that so many missed.
00:10:20
Deep seek proved this new scaling law,
the test time scaling or long think
00:10:25
law, it requires, just more accelerated
computing infrastructure to be installed.
00:10:32
So in a nutshell, Pre training you can
think of, if AI is like a student, pre
00:10:37
training is like going through undergrad
studies, post training is like going
00:10:41
through a specialized program, maybe like
a doctorate or an MBA program, and then
00:10:47
that long thinking is like maybe taking
like a long walk to think about a problem
00:10:52
or taking time to meditate on something.
00:10:55
And when we get to that last
scaling law, this is where NVIDIA
00:10:59
has been trying to make the point
that DeepSeek is validating.
00:11:02
It doesn't put them at risk because this
type of very, very deep reasoning that
00:11:09
these AI models are now only just at
the beginning of being able to pull off.
00:11:14
It takes sometimes upwards of 100x more
computing power to make it all possible.
00:11:20
So if you think more software is
going to have more AI features
00:11:25
embedded into it in the coming years.
00:11:28
This is all great news
for a company like NVIDIA.
00:11:31
Yes, we have the gross margin issues
to worry about right now, this year.
00:11:36
Revenue is not going to continue growing
at a 100 percent plus rate year over year.
00:11:42
So slowing, or maybe we should just
say normalizing revenue growth, a
00:11:47
bit of a decline in gross margin.
00:11:50
But, it still looks like we are very much
in the early days still of this whole
00:11:55
accelerated computing and A. I. Industry.
00:11:57
Let's round out this discussion
with point number three.
00:12:00
When is it time to sell
your NVIDIA position?
00:12:04
We see three important things
to keep an eye out for.
00:12:07
The first one, big tech capital
expenditures, flatlining or declining.
00:12:11
Currently, as you probably know,
that is definitely not happening.
00:12:16
Capital expenditures continue to ramp
up for all of the big tech spenders.
00:12:20
The second point Nvidia's,
revenue and guidance flatlining.
00:12:24
We certainly haven't seen
that happening yet either.
00:12:27
And the third point valuation
looks absurdly high.
00:12:31
So does the valuation
look absurdly high yet?
00:12:33
Well, I mean, this has been the argument
for two years now, if not longer, if
00:12:37
you include during the bear market where
the argument was still Nvidia stock is
00:12:42
it's still too expensive, even though
we were in a bear market and they had no
00:12:46
earnings, of course it looked expensive.
00:12:48
But you know, here we are years
later, still having this conversation.
00:12:52
And at least as far as we're concerned
and what our portfolio looks like and what
00:12:57
its needs are, there's nothing out of the
norm or out of the ordinary with NVIDIA's
00:13:02
valuation that would make us want to
sell any of it at this particular point.
00:13:08
So we just plugged in a few simple numbers
into a reverse DCF estimate reverse DCF.
00:13:16
Remember we are just simply trying to
come up with some sort of expectation
00:13:21
that gets us to fair value on what
the stock price is right now, which
00:13:26
is about a hundred and thirty dollars
after hours after the q4 earnings call.
00:13:32
We used full fiscal year 2025
GAAP earnings per share of 2.
00:13:38
94. Earnings per share compound
annual growth rate of 30
00:13:42
percent for the next two years,
00:13:44
the end of fiscal 2027, which will be
the 12 months ending in January, 2027.
00:13:50
Terminal growth rate of 6.
00:13:52
5%, thereafter discount rate of
10 percent gets us to around 130,
00:13:58
which is the stock price currently.
00:14:01
Is that fair?
00:14:03
Yeah.
00:14:03
I mean, that's the question.
00:14:04
If you like using a reverse DCF estimate
to try to gauge whether a stock is a
00:14:12
good buy or, worth holding still, or
things have gotten completely absurd
00:14:17
and, and out of, out of whack, out of
alignment, this can be one good tool
00:14:23
to help you make that assessment.
00:14:26
So that's what we've done here, and
looking at these numbers, we actually
00:14:29
think things are looking more reasonable
than ever, actually, for NVIDIA, and so
00:14:34
we see no reason to sell our position.
00:14:38
Now, that doesn't mean we're going to be
buying, that's just us and for our needs.
00:14:41
We don't need any more NVIDIA
stock in our portfolio.
00:14:44
If you've been with us over on Semi
Insider, you know that we have been
00:14:48
putting in a little bit extra work,
especially in the last six months or so
00:14:53
to diversify our portfolio a bit more and
raise our cash position just a little bit.
00:14:59
So that is still where
we're personally at.
00:15:03
That's our look on Nvidia stock.
00:15:04
We think it's going to be a great year
in spite of there being some short
00:15:09
term growing pains, but everything
looks good as far as we can tell.
00:15:15
Thanks everybody for watching this
episode of Chip Stock Investor.
00:15:19
Make sure you sub to the channel, check
out Semiconductor Insider membership
00:15:23
link in the video description below.
00:15:25
Just 10 a month that gets you access
to Discord server, where we have all
00:15:29
of our video notes, live Q and A's,
which we did one right after NVIDIA's
00:15:34
earnings, lots of great benefits.
00:15:36
Check it out.
00:15:36
We'll see you again soon
at Chip Stock Investor.
00:15:39
Trying to have a
productive record session.
00:15:42
Rhea's also excited about NVIDIA.
00:15:46
Okay.
00:15:47
You're gonna do your,
uh, your high energy?
00:15:49
Me do my high energy.
